How do I install the Agent Browser skill?

Run npx skillmds@latest add getsentry/agent-browser in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Agent Browser safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Agent Browser?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Agent Browser free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Agent Browser?

Sentry (@getsentry) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Agent Browser by getsentry — Agent Skill

Name: Agent Browser
Author: Sentry

Agent Browser

Automate websites with the agent-browser CLI for navigation, form flows, screenshots, extraction, and repeatable browser tasks. Use when users ask to interact with web apps, gather evidence from pages, or run browser-based QA steps.

Sentry 59d6046 11 files · 44.9 KB Updated Sep 17, 2026 845 repo stars

File contents

Browser Automation

Use this skill when work requires real browser interaction instead of API-only lookups.

Core Workflow

Open target page:

agent-browser open <url>

Capture interactive refs:

agent-browser snapshot -i

Interact by ref:

agent-browser click @e1
agent-browser fill @e2 "value"

Re-snapshot after every navigation or significant DOM change:

agent-browser snapshot -i

Capture result evidence:

agent-browser screenshot --annotate <file.png>

Operational Rules

Use agent-browser directly; do not use npx agent-browser.
Prefer one named session for multi-step flows:
- agent-browser --session <name> ...
Use chained commands with && only when no intermediate output is needed.
For commands that need refs (@e*), run snapshot -i first and use fresh refs.
If page load is async-heavy, wait explicitly:
- agent-browser wait --load networkidle
For user-facing findings, include screenshots and the exact page URL.
When a user asks to see/share a screenshot in Slack, use sendFiles with the screenshot path in files so the image is included in the active conversation.
Never claim a screenshot was shared unless sendFiles returned success in this turn.
If sendFiles is unavailable or file attachment fails, state the failure clearly and include the saved screenshot path or error instead of claiming success.

Common Patterns

Form flow

agent-browser --session demo open https://example.com/signup
agent-browser --session demo wait --load networkidle
agent-browser --session demo snapshot -i
agent-browser --session demo fill @e1 "user@example.com"
agent-browser --session demo click @e2
agent-browser --session demo wait --load networkidle
agent-browser --session demo screenshot --annotate /tmp/signup-result.png

Structured extraction

agent-browser open https://example.com
agent-browser snapshot --json > /tmp/page.json
agent-browser get text body > /tmp/page.txt

Reference Files

Read only what is needed:

Command surface: references/commands.md
Auth vault + login flows: references/authentication.md
Session/state persistence: references/session-management.md
Snapshot refs and stale-ref handling: references/snapshot-refs.md
Proxy/network setup: references/proxy-support.md
Recording and video capture: references/video-recording.md
Performance profiling: references/profiling.md

Templates

Authenticated flow bootstrap: templates/authenticated-session.sh
Capture workflow skeleton: templates/capture-workflow.sh
Form automation skeleton: templates/form-automation.sh

---name: agent-browserdescription: Automate websites with the agent-browser CLI for navigation, form flows, screenshots, extraction, and repeatable browser tasks. Use when users ask to interact with web apps, gather evidence from pages, or run browser-based QA steps.---# Browser AutomationUse this skill when work requires real browser interaction instead of API-only lookups.## Core Workflow1. Open target page:- `agent-browser open <url>`2. Capture interactive refs:- `agent-browser snapshot -i`3. Interact by ref:- `agent-browser click @e1`- `agent-browser fill @e2 "value"`4. Re-snapshot after every navigation or significant DOM change:- `agent-browser snapshot -i`5. Capture result evidence:- `agent-browser screenshot --annotate <file.png>`## Operational Rules- Use `agent-browser` directly; do not use `npx agent-browser`.- Prefer one named session for multi-step flows: - `agent-browser --session <name> ...`- Use chained commands with `&&` only when no intermediate output is needed.- For commands that need refs (`@e*`), run `snapshot -i` first and use fresh refs.- If page load is async-heavy, wait explicitly: - `agent-browser wait --load networkidle`- For user-facing findings, include screenshots and the exact page URL.- When a user asks to see/share a screenshot in Slack, use `sendFiles` with the screenshot path in `files` so the image is included in the active conversation.- Never claim a screenshot was shared unless `sendFiles` returned success in this turn.- If `sendFiles` is unavailable or file attachment fails, state the failure clearly and include the saved screenshot path or error instead of claiming success.## Common Patterns### Form flow```bashagent-browser --session demo open https://example.com/signupagent-browser --session demo wait --load networkidleagent-browser --session demo snapshot -iagent-browser --session demo fill @e1 "user@example.com"agent-browser --session demo click @e2agent-browser --session demo wait --load networkidleagent-browser --session demo screenshot --annotate /tmp/signup-result.png```### Structured extraction```bashagent-browser open https://example.comagent-browser snapshot --json > /tmp/page.jsonagent-browser get text body > /tmp/page.txt```## Reference FilesRead only what is needed:- Command surface: [references/commands.md](references/commands.md)- Auth vault + login flows: [references/authentication.md](references/authentication.md)- Session/state persistence: [references/session-management.md](references/session-management.md)- Snapshot refs and stale-ref handling: [references/snapshot-refs.md](references/snapshot-refs.md)- Proxy/network setup: [references/proxy-support.md](references/proxy-support.md)- Recording and video capture: [references/video-recording.md](references/video-recording.md)- Performance profiling: [references/profiling.md](references/profiling.md)## Templates- Authenticated flow bootstrap: [templates/authenticated-session.sh](templates/authenticated-session.sh)- Capture workflow skeleton: [templates/capture-workflow.sh](templates/capture-workflow.sh)- Form automation skeleton: [templates/form-automation.sh](templates/form-automation.sh)

Frequently asked questions